Estimating the three-dimensional structure of canopy foliage based on the light measurements in a Betula ermanii stand

Abstract

In order to more accurately evaluate the functional activity of forest stands by canopy production and evapotranspiration, we improved the methods for field measurements and statistical modeling to estimate foliage configuration (spatial distribution of leaves) while simultaneously reconstructing the three-dimensional photosynthetically active photon flux density (PPFD) distribution (PPFD pattern) in a forest canopy. By using a sensor (photodiode) array, a PPFD pattern was observed in summer 2002 under the canopy in an even-aged, pure stand of Japanese mountain birch Betula ermanii Cham. (17-years old) in Hokkaido, northern Japan. A Markov chain Monte Carlo (MCMC) sampling technique is developed such that a set of foliage configurations generated by the model referred to as the Gibbs foliage canopy (GFC) approximates the field-measured PPFD pattern. The posterior distribution of the foliage configurations is generated by the parallel tempering MCMC of eight independent series of foliage configurations. The GFC model generated the posterior distribution of the LAI estimates (mean 4.56) that appeared to be appropriate in comparison to other LAI estimates of the B. ermanii stand based on the indirect and nondestructive methods by LAI-2000 (LAI = 3.43) and litterfall traps (LAI = 5.56) because they could be under- and overestimated, respectively. Our evaluations of the canopy production and evapotranspiration rates suggest that the relationship between LAI and canopy functions was not very simple because it depended on the nonlinear functional forms of the leaf responses of photosynthesis and transpiration to PPFD. The current study demonstrates an application of MCMC techniques that can generate a set of possible structures of unobserved/unobservable objects based on the high-resolution dataset obtained by some indirect (or remote-sensing) methods.